Note: Descriptions are shown in the official language in which they were submitted.
CA 02454983 2004-O1-07
SYSTEM FOR HIGH-SPEED APPLICATIONS OVER SERIAL MULTI-DROP
COMMUNICATION NETWORKS
The present invention relates to a system and a methodology that enable TDM
(Time
Division Multiplexing) applications over an adapted EIA-485 network.
EIA-485 (formerly "RS-485") is a standard serial hardware protocol for mufti-
drop
communication networks that specifies up to 32 drivers and 32 receivers on a
single (2-
wire) bus. Maximum data rates are 10 Mbps at 1.2 m or 100 Kbps at 1200 m.
Today, some manufacturers are providing EIA-485 transceivers with pre-emphasis
and
corresponding receiver de-emphasis to double the distance at data rates over
400 kbps.
The same devices may be used to increase the data rate for a specific distance
(up to 35
Mbps for distances less than 10 m) and to allow up to 128 transceivers on the
bus.
EIA-485 is one of the most used mufti-drop communication protocols and one of
the
most economic physical layer protocols. Many electronic devices encompass EIA-
485
ports. It is also widely used for electronic systems on-board transit
vehicles.
Of interest are the following documents: "Electrical Characteristics of
Generators and
Receivers for Use in Balanced Digital Multipoint Systems (ANSI/TIA/EIA-485-A-
98)(R2003)"; "Comparing Bus Solutions, Application Report SLLA067 - March
2000",
Texas Instruments, http://polima~e.polito.it/~lava~no/esd/bus.pdf.
The present invention defines the protocols in the physical layer and the
protocols above
the physical layer to allow the sending of Time Division Multiplexing (TDM)
based
digital data signals over an EIA-485 bus at a rate up to 10 Mbps at a distance
exceeding
150 m.
The present invention encompasses an end-to-end system that broadcasts several
audio
stereo channels with hi-fi quality to individual passenger seats on-board a
public transit
vehicle. The system according to the invention comprises an audio encoder, an
audio
decoder and a data repeater. Each passenger can choose the audio channel that
he (she)
wants to hear for his (her) entertainment.
The same system can be used to provide other applications.
The same TDM multiplexing principle may be applied to other communication
protocols,
for example RS-422 to enable other applications.
In a possible embodiment, the uncompressed audio data is 16 bits per channel
and each
channel is sampled at 43.17 KHz. For a synchronization purpose, the header
size is 17
bits. A parity bit is preferably included. This is illustrated in Figure 1.
Figure 2 shows an end-to-end network architecture for an in-seat audio
entertainment
system.
CA 02454983 2004-O1-07
Figure 3 presents an architecture of an audio encoder, which is also called
audio server
hereinafter. Several audio channels are converted to digital, multiplexed with
TDM and
transmitted over an EIA-485 network.
Figure 4 presents an architecture of an audio decoder. The decoding occurs at
the
passengers' seats.
Figure 5 shows an architecture of a data repeater. This component is optional.
It is used to
clean and to regenerate the signal when it is carned over long distances.
There is no commercially available implementation of several data channels
multiplexed
in TDM format and sent over a single EIA-485 bus.
To develop the end-to-end system, a combination of commercially available
devices and
innovative technologies is utilized.
The commercially available devices utilized are the Delta-Sigma ADC and DAC
converters for the analog to digital conversion and vice-versa. Another off
the-shelf
technology that is used is the EIA-485 transceivers with pre-emphasis (and the
corresponding receivers with de-emphasis), which extends the distance and
increases the
data rate of reliable communication by reducing intersymbol interference (ISI)
caused by
long cables. These transceivers are programmable for data rates up to 10 Mbps
and they
allow up to 128 transceivers on the bus.
EIA-485 is a physical layer protocol. Different manufacturers are implementing
different
packet formats for the data layer over an EIA-485 bus. Simple ASCII commands
are
often provided. Typically, the minimum overhead is 32 bits (4 bytes), which is
not
optimal for some applications; this overhead is reduced in the present
invention to
optimize bandwidth usage.
E1/T1 mufti-channel broadcasting devices are commercially available. E1
supports a 2
Mbps rate and T1 supports 1.5 Mbps, whereas the improved EIA-485 devices
support up
to 35 Mbps. Therefore, for the same sampling frequency and quality (44.1 kHz
for hi-fi
stereo quality), less channels would be supported by an E1/T1 solution.
Furthermore,
E1/T1 is a point-to-point solution. To adapt it to a mufti-drop network,
additional devices
would be required, increasing complexity and costs.
The present invention solves several problems of the art. It provides an end-
to-end system
that allows the sending of multiple channels of digital data over a single EIA-
485 bus. It
reduces the amount of bandwidth required for applications over EIA-485. It
allows the
deployment of high-speed applications over longer distances. It reduces the
number of
buses required, and related equipment. It avoids the usage of more expensive
mufti-drop,
mufti-point communication protocols. A very simple synchronization technique
between
the encoder, the decoder and the repeater is developed. A method to correct
corrupted
data is also provided. Logarithmic compression is proposed to further increase
the
2
CA 02454983 2004-O1-07
number of channels supported on a single EIA-485 bus. The invention takes very
little
space from both cabling and equipment perspectives. Space economy is a big
advantage
for deployment in transit vehicles.
The main driver behind the invention is cost-effectiveness. In order to reduce
the overall
cost of an in-seat audio entertainment project, a modified, enhanced EIA-485
bus was
selected as the digital data link. EIA-485 has three major advantages: it's a
bus, it's
multi-drop and it's an economic technology for the application. The challenge
was to
send 5 uncompressed digital stereo hi-fi channels on this 10 Mbps limited
channel. The
audio by itself required 7.056 Mbps. The more the data rate is increased, the
more
potential error may occur.
A detailed description of the entire system is given hereinafter.
Here's an overview of the end-to-end solution and the key technologies used.
~ Analog to digital converters and vice-versa. Delta-Sigma ADC and DAC
converters are used for this purpose. This type of converters provide i) over
sampling
and ii) closed-loop modulation. This combination results in a higher quality
digital
signal. This is an off the-shelf technology.
~ EIA-485/RS422 transceiver with pre-emphasis. A commercially available EIA-
485/RS422 transceiver with pre-emphasis was used. With EIA-485 networks, the
trade-off has always been less distance at a higher rate or greater distance
at a lower
rate. Pre-emphasis extends the distance, and increases the data rate of
reliable
communication by reducing the intersymbol interference caused by long wires.
~ Data frame format and protocol. A novel data frame format and protocol is
provided to multiplex TDM signals over EIA-485 networks with a header of
minimal
size and error correction is included. The method according to the invention
reduces
the EIA-485 overhead to 18 bits in uncompressed format, which is lower than
any
commercially available implementation. Five separate channels are multiplexed
in
TDM format and sent over a single EIA-485 bus. Each uncompressed stereo
channel
contains 32 bits. Therefore, up to 6 simultaneous uncompressed channels (up to
10
compressed channels) can be supported by a single EIA-485 bus because the
total
bandwidth is less than 10 Mbps. It is always an advantage to consume the least
amount of bandwidth possible, because the more the 10 Mbps limit is
approached, the
higher the probability of potential errors.
~ Synchronization between the different components of the system. A
synchronization method between the encoder, the decoder and the repeater is
provided. There's a crystal oscillator in each device. The synchronization
method
doesn't require any adjustments. It locks the phase between the local clock
and the
external clock. The external clock is found within the incoming data stream.
The
phase comparison always occurs after at least four consecutive high bits, this
will
give always the same threshold for the phase comparator. The phase is then
locked
CA 02454983 2004-O1-07
for the next four-bit pattern. The header has at least four consecutive high
bits to
guarantee that the phase comparison and the phase lock are performed at least
once
for every message.
Error management method. An error management method, that in the audio
application allows the end-user to hear no audible difference in case of
errors, is also
provided. For any communication network, there's some data corruption
probability.
In order to manage such potential problem, a parity bit is provided to check
the data
integrity. The parity is part of each frame of 178 bits; it should be enough
to keep a
good and reliable audio quality. If an error occurs, the analyzer won't be
able to select
the wrong data position, so the error handling strategy is to send the
previous audio
data again. The same strategy is valid if the frame detection is lost or the
phase locked
is missing. The last two events should be exceptional.
Logarithmic compression. Logarithmic compression can increase the number of
data channels that can be transmitted over a single EIA-485 data link. A
logarithmic
encoder/decoder can be employed in order to use a greater portion of the
available
levels for weak signals. This process can be thought of as compressing the
signal in
amplitude. Using this technique, it is possible to achieve up to 10 audio
stereo
channels on a single EIA-485 data link.
~ Network repeater. Even if its functionality seems to be very simple, this
module is
complex. It uses the same kind of technology developed for the audio decoder
for the
data tracking but the emphasis has been focused on the high tracking precision
instead of highly reactive tracking. The repeater task is to reduce jittering,
wandering
and keep data integrity.
EIA-485 is one of the most economic serial multi-drop network solutions.
Reduced
amount of cabling and peripheral devices also provides significant cost
savings.
The following provides a detailed description of the invention, the hardware
architecture
of the system according to the invention and relation with its related
peripherals. The
following abbreviations are used hereinafter:
CD Compact Disc
ISI Intersymbof Interferenca
LUT Look Up Table
MBPS MegaBits Per Second
MTBF Mean Time Between
Failure
PLL Phase Locked Loop
PPM Parts Per Million
VCXO Voltage Controlled crystal (X) Oscillator
The entire electronic audio world has changed since the event of the CD player
birth.
Huge technical problems had to be overcome in order to give consumer a small,
reliable
and affordable product. The most challenging problems were optical, high
quality audio
4
CA 02454983 2004-O1-07
and power consumption. In order to achieve the hi-fi audio, the engineers
looked for a
more digital than analog solution. This technology is now available on single
chip
solution and is called Delta-Sigma converters. This technology is based on the
over
sampling concept.
The present invention relates to a digital audio network that can transport up
to 5 stereo
audio streams (10 in phase 2), involving streaming. The transport medium
selected is a
modified RS~185 that includes pre-emphasis in order to achieve higher data
rates with
more loads.
The complete architecture is shown in figure 2. The first stage of the system
is the Audio
Server that makes the data ADC conversion and framing. Up to five uncompressed
or ten
compressed stereo channels can be sent on one single RS-485 channel. This
section is
located on a 19 inches rack mount facility. The next stage is the Audio
Decoder, which is
in the armrest of the user seat. The last stage is the Data Repeater, which
rebuilds,
cleans up and repeats the RS-485 signal to the next car.
The proposed architecture is based on four different technologies borrowed
from
different electronic fields. The first technology is the Sigma-Delta ADC and
DAC
converters.
Delta-Sigma ADC converters differ from other ADC approaches by sampling the
input
signals at a much higher rate than the maximum input frequency. Traditional,
non-over
sampling converters such as successive approximation ADCs perform a complete
conversion with only one sample of the input signal. Another unique
characteristic of the
Delta-Sigma converter method is that a closed loop modulator is used. The
modulator not
only continuously integrates the error between a crude ADC and the input
signal, but also
attenuates noise. This combination of over sampling and closed-loop modulation
creates
a very powerful technique.
Delta-Sigma concept can also be applied on DAC converters. The main difference
between Delta-Sigma ADC and DAC lies in the rate of the output signal. In the
ADC
section, decimation is used to reduce the high frequency low-resolution pulses
to lower
frequency, higher resolution words. Delta-Sigma DACs, on the other hand, do
the
reverse. Here a process called interpolation is performed that samples the
digital outputs
at a higher rate. This produces a high-resolution/frequency output that is
easily low pass
filtered for an analog output.
The second technology is the RS-485/RS422 transceiver with pre-emphasis, which
extends the distance, and increases the data rate of reliable communication by
reducing
the intersymbol caused by long wires. The pre-emphasis drivers incorporate
four voltage
levels (strong high, strong low, normal high, normal low). Pre-emphasis is
necessary only
when the data pattern changes and not during the intervals when the voltage
remains at
the same logic level.
CA 02454983 2004-O1-07
The third technology is the fitter attenuation based on analog and/or digital
phase locked
loop. Signal fitter is primarily due to intersymbol interference (ISI). ISI is
the net effect of
several causes of signal degradation. One cause is the attenuation and the
dispersal of
frequency components that result from signal propagation down a transmission
line.
Another cause is the variation of rise and fall times that follows the varying
sequences of
one and zero known as "pattern-dependent skew". A data pulse responds to these
effects
with a loss of amplitude, displacement in time, rounded edges, and a
"smearing" of the
pulse into adjacent time slots, or unit intervals. By locking the phase of our
repeater on
one particular, intentionally generated, clean pulse, the rest of the audio
data stream can
be sampled based on this phase and therefore most of the fitter can be
removed.
The fourth technology is the logarithmic compression. A logarithmic
encoder/decoder is
employed in order to use a greater portion of the available levels for weak
signals. This
process can be thought of as compressing the signal in amplitude. Using this
technique,
up to 10 audio stereo channels on a single RS-485 data link may be achieved.
Figure 1 shows the audio data frame structure. The uncompressed audio data is
I6 bits
per channel; each channel is sampled at 43.17 KH2. For a synchronization
purpose, the
header size is 1? bits. One parity bit is also included.
The necessary bandwidth for a 5 uncompressed stereo channel is then:
((5 channels x 2 (stereo) x 16 bits) + 17 bits (header) + 1 parity bit) x
43.17 KHz =
7.6843 Mbps
The necessary bandwidth for a 10 compressed (16 to 10) stereo channels is:
((10 channels x 2 (stereo) x 10 bits) + 11 bits (header) + 1 parity bit) x
43.17 KHz =
9.152 Mbps
In order to reduce the cost of the overall system, it has been decided to
blind broadcast
the signal through the cars. A simple but powerful error remodeling is
implemented at the
decoder level. The repeater task will be to provide a clean error free, fitter
attenuated
signal for the next car.
A common power source for all the system components is provided. The selected
secondary isolated voltage is 12 Volts. 12 Volts is the basic system voltage.
Each
equipment should provide its own isolated power supply. The primary voltage
can be the
usual ones. Each system component has an RS-485 enhanced serial port. In order
to
reduce any ground looping and to stay within the standard limits, all
communication
interfaces are isolated.
The audio server is digitizing five analog stereo channels. In order to be
able to reach hi-
fi levels, Sigma-Delta converters are used. The analog interface between audio
servers
and Sigma-Delta converters is composed of an amplifier and a low pass filter.
The
6
CA 02454983 2004-O1-07
amplifier is provided to match the conversion level and the low pass filter to
reduce the
possible aliasing distortion. LUT compression can be included if desired.
In order to reduce the cost of the system, a modified, enhanced RS-485 has
been selected
as the digital data link. The challenge is to send 5 uncompressed digital
stereo hi-fi
channels on this 10 Mbps limited channel. The audio by itself requires 6.907
Mbps. More
the data rate increases, more fitter and potential error occur. The overhead
is reduced by
just adding a 17-bit header and one parity bit. If in somewhere in the data
stream, the
same pattern as the header occurs, the actual audio channel is just replaced
by 0, the
decoder is designed to manage this information.
The design of the header provides immunity against any problems. The header
has less
transition then audio data in order to give a very good eye pattern at the
line level. By
using one parity bit for a complete audio frame, a very effective way to
detect error in
each frame should be achieved.
Different frequencies are necessary to produce the audio data stream at the
right bit rate.
Actually the frequency relation between the output bit rate and the reference
frequency is
0.6953125 (89/128). This relation comes from the amount of channels sent and
the
overhead added to the data versus the sampling rate of the over sampling
converter. To
generate such frequency, a phase-locked loop with one programmable divider in
the
feedback loop (I~ and one at the output (M) is needed. The frequency generated
will then
be Fout = Fref * N1M.
A way to mute all the digital audio channels may be provided. The command may
come
from an external device. A way this feature may be implemented is to continue
sending
the previous data as long as the mute command is active.
It is very hard to test an audio system with real data. Audio data is complex
and following
its path through the system would be a complex task. In order to ease the
testing of the
complete system, it has been opted to put a pattern generator in the audio
server. Instead
of getting samples from the ADC converters, some predefined patterns from this
device
are obtained.
The audio decoder is the most challenging part of the system. Even if it is
complex, it
remains inexpensive. The decoder has the capability to give hi-fi sound
quality even in
harsh environment. The complete system has been designed to provide good
quality
sound without any feedback to the server. The decoder synchronizes to incoming
data
stream, extracts the stereo channels, checks if there's a possible data
corruption and takes
action. Even if corruption happens, the listener does not hear any glitch.
The basic frequency is the same between the server and the decoder. Even if
they're
crystal based, oscillators have some part per million of frequency variations.
The strategy
here is to sample the data in at 8 times the incoming data rate and test the
phase relation
with the internal reference. The algorithm enables the reference to have some
phase
variations. With time, this variation will become unacceptable and the phase
correction
7
CA 02454983 2004-O1-07
will then occur. Looking to the figure 4, a feedback from the sampler stage to
the VCXO
can be seen. Even if this oscillator is crystal based, its frequency can be
varied up to ~
100 ppm. Using this feature, it is possible to avoid any slip buffer to happen
because our
frequency reference will always be locked on the incoming data stream. The
phase
comparison always occurs after at least four consecutive high bits, this will
give always
the same threshold for the phase comparator. The phase is then locked for the
next four-
bit pattern (the header has such pattern).
The architecture shown in figure 4 does not reflect the decompression
function. This
optional feature can be based on the following simple principle: an
antilogarithm function
is encoded into a dedicated memory. This function is the reversal of the
server one.
Having this, a good signal dynamic with a minimum distortion can be obtained.
The user interface may be limited to four push buttons. Each switch is de-
bounced and
the corresponding command is sent to the right stage. The volume control is
made using
logarithmic digital potentiometers receiving their value through a serial
port. Each
potentiometer has its own counter and shift register. For the channel
selection, all the
design process is controlled and a simple counter for each channel is
provided.
For any communication network, there is some data corruption probability. In
order to
manage such potential problem, a parity bit is provided to check the data
integrity. The
parity is part of each frame of 178 bits, which should be enough to keep a
good and
reliable audio quality. If an error occurs, the analyzer won't be able to
select the wrong
data position. The error handling strategy is to send the previous audio data
again. The
same strategy is valid if the frame detection is lost or the phase locked is
missing. The
last two events should be exceptional.
The configuration of a regular digital audio network is to have one audio
server for up to
hundreds of audio decoders. This high volume production requires the design to
be quick
to check by manufacturing staff. The decoder design surrenders the user's
control to an
external intelligent device. This device will send volume and channels up/down
command and will check for audio harmonic distortion and communication
reliability.
This device will also have the task to check if the push button interface is
o.k. All the
production tests will use a 3 wires interface: data; clock; load.
There is no adjustment for the audio decoder required. The only thing to set
in the field is
putting the strap for the RS-485 terminals if necessary.
Even if the functionality of the data repeater seems to be very simple, this
module is
complex. It uses the same kind of technology developed for the audio decoder
for the
data tracking but the emphasis has been focused on the high tracking precision
instead of
highly reactive tracking. The repeater task is to reduce jittering, wandering
and keep data
integrity.
As mentioned earlier, the strategy for phase tracking is similar as for the
audio decoder.
The big difference is its great accuracy. Instead of sampling the incoming
data at 8 times
CA 02454983 2004-O1-07
its frequency, it is sampled at 16 times. More, instead of having a feedback
with tree
different conditions (+100ppm, Oppm, -100ppm), a 16 levels feedback to the
VCXO is
added. The comparison could happen once per frame only and the phase would be
locked
for the rest of the data stream. The event that triggers the phase locking is
a particular
pattern found in the header. This header reduces the electrical effects
relative to the
communication cable.
Transmission lines are not immune against electrical spikes or high-energy
radiation
burst. In order to reduce data corruption at the repeater level, some digital
filtering on the
incoming data may be included. Sampling the incoming data many times in its
valid
period of time and making a correlation between them does this filtering.
The repeater is the last equipment along the transmission line inside one car.
It should
include RS-485 terminals.
The following provides the electrical specifications of the system:
Audio Server:
Standards TBD
Resolution 16 bits
Audio bandwidth 20 KHz
Harmonic distortion TBD
Galvanic isolation power1500V
Galvanic isolation RS-4851500V
Stereo channels 5 phase 1, expected
10 phase 2
Audio source 1 Vpp into 600 Ohms
Temperature -40 to 85 Celsius
MTBF TBD
Audio decoder:
Standards TBD
Resolution 16 bits
Audio bandwidth 20 KHz
Harmonic distortion TBD
Galvanic isolation power1500V
Galvanic isolation RS-4851500V
Stereo channels 5 phase 1, expected
10 phase 2
Audio out TBD
Temperature -40 to 85 Celsius
MTBF TBD
9
CA 02454983 2004-O1-07
Repeater:
Standards TBD
Jitter/wander attenuationTBD
Galvanic isolation power1500V
Galvanic isolation RS-4851500V
Temperature -40 to 85 Celsius
MTBF TBD
It should be understood that changes and modifications may be made in the
above
embodiments without departing from the essence of the invention.
